Bayesian probability transforms for BM25 retrieval scores

Random stuff I've been working on

Local Temperature Scaling for Probability Calibration

Non-Parametric Calibration for Classification (AISTATS 2020)

UQ4DD: Uncertainty Quantification for Drug Discovery

This repository implements Pozzolo, et al., (2015)'s probability calibration for imbalanced data.

Decision-first fraud screening: Streamlit analytics UI + FastAPI inference, artifact-locked RF/XGB models, and threshold policy controls.

No description provided.

khalib is a classifier calibration algorithm based on Khiops

Cost-aware credit card fraud detection pipeline: time-based split, probability calibration, and business-aligned threshold tuning (AUPRC-first).

A simple yet effective post-processing method for detecting unknown intent in dialogue systems based on pre-trained deep neural network classifiers

My Master's thesis on Bayesian Classification with Regularized Gaussian Models

Decision-grade donor outreach policy: calibrated P(donated_next_6m) scoring + budgeted Top-K actions + net-benefit optimal threshold, with exported deployable artifacts.

Bayesian probability transforms for BM25 retrieval scores (TypeScript)

Interpretable credit risk modeling using real-world lending data, with emphasis on probability calibration, decision relevance, and scalable machine learning workflows.

No description provided.

Code for the internship report. Sample × Category Probability Calibration in Two Dimensions.

🧠COGNITIVA-AI: IA intermodal (clínica+MRI) para cribado temprano de Alzheimer; probabilidades calibradas, umbrales por cohorte (S2) y release reproducible.

Machine learning–based cardiovascular risk screening app focused on high recall

End-to-end credit risk modeling pipeline with probability calibration, risk bucketing, and SHAP explainability.

Code and experiments for the preprint: "Bayesian Neural Network Versus Ex-Post Calibration For Capturing Prediction Uncertainty".

End-to-End Machine Learning pipeline for sports match prediction. Features advanced feature engineering (rolling windows, dynamic Elo), temporal validation, and probability calibration using LightGBM.

Develop interpretable credit risk models using lending data to improve default probability estimates for sound financial decisions.